Hello,

I am curious if anyone knows of a good justification for including WM in the default analysis mask? I know of at least one publication (Gawryluk JR, Mazerolle EL, D'Arcy RCN, 2014) which argues that WM should be included in more analyses. However, I still have a number of questions:

1. If WM is included, should the default denoising step still regress out WM signals as a nuisance covariate? Would CSF signals be sufficient to capture motion and other artifacts?

2. If WM is included, should the GM-WM interface be avoided to reduce partial volume effects?

3. Are there other benefits of including WM? For example, it seems possible that activations could extend into the WM, and a GM-WM mask might capture this cluster, whereas a GM-only mask might cut off the WM section, preventing it from passing threshold.

Any advice appreciated,
Elliot